Automatic Trend Change Detection in n_changepoints

[bikefelix]

It is my understanding that the number of trend changes should be automatically determined by NeuralProphet (default settings), with a maximum limit set by the value of n_changepoints. However, in practice, I consistently observe that the number of trend changes exactly matches the number I specify. For example, when I set n_changepoints = 50, exactly 50 trend changes are used.

Similarly, when I replicate this specific code, exactly 10 trend changes are selected, despite different results in the original notebook.

I'm puzzled by this outcome as I expected the algorithm to potentially select fewer trend changes than the maximum specified. Could someone please clarify if this is the intended behavior or if I might be overlooking something?